Google DeepMind Introduces Gemini 4 Argon for Robotics, Showing Robots Handling Real Objects

At a glance

Google DeepMind has introduced Gemini 4 Argon, an AI model used for robot tasks. The model's ability to understand the position and movement of objects and to generate robot motion code has improved significantly, according to the company. The article includes examples of robots stepping on a trash can pedal and standing up a cup, among other tasks.

Key points

1. 1

   The model scored 87% on the ERQA spatial reasoning test, higher than other leading models, according to the company.

2. 2

   The article gives examples of a four-legged robot and an arm robot learning to handle real objects.

3. 3

   It also showed a method for transferring hand movements from videos of people to robot hands.

4. 4

   The article contains no information about a commercial product launch outside the lab or about pricing.

Why it matters

For robots to handle many objects on their own in homes or factories, they need this kind of physical understanding.